Market structure, scale economies, and industry

Transcription

Market structure, scale economies, and industry
discussion papers
FS IV 00 – 08
Market Structure, Scale Economies,
and Industry Performance
Rabah Amir
University of Southern Denmark at Odense
August 2000
ISSN Nr. 0722 - 6748
Forschungsschwerpunkt
Marktprozeß und Unternehmensentwicklung
Research Area
Market Processes and
Corporate Development
Zitierweise/Citation:
Rabah Amir, Market Structure, Scale Economies, and Industry
Performance, Discussion Paper FS IV 00-08, Wissenschaftszentrum
Berlin, 2000.
Wissenschaftszentrum Berlin für Sozialforschung gGmbH,
Reichpietschufer 50, 10785 Berlin, Tel. (030) 2 54 91 - 0
ABSTRACT
Market Structure, Scale Economies, and Industry Performance
by Rabah Amir
We provide an extensive and general investigation of the effects on industry
performance (profits and social welfare) of exogenously changing the number of firms
in a Cournot framework. This amounts to an in-depth exploration of the well-known
trade-off between competition and production efficiency. We establish that under scale
economies, welfare is maximized by a finite number of firms. Our results shed light on
several theoretical issues and policy debates in industrial organization, including the
relationship between the Herfindahl index and social welfare, free versus socially
optimal entry, concentration and profitability, destructive competition and natural
monopoly. Our analytical approach combines simplicity with generality.
Keywords: Cournot oligopoly, returns to scale, entry, equilibrium comparative statics
JEL codes: D43, D60, L13, L40
ZUSAMMENFASSUNG
Marktstruktur, Größenvorteile und Marktergebnis
In dem Beitrag wird im Rahmen eines Cournot-Ansatzes eine allgemeine Untersuchung
der Auswirkungen der exogen sich ändernden Anzahl von Unternehmen auf das
Marktergebnis (Gewinne und soziale Wohlfahrt) vorgestellt. Dabei wird in einer analytischen Betrachtung das bekannte Trade-off zwischen Wettbewerb und Produktionseffizienz herausgearbeitet. Es wird gezeigt, daß beim Vorliegen von Größenvorteilen
die Wohlfahrt durch eine endliche Zahl von Firmen maximiert wird. Das Ergebnis
beleuchtet verschiedene Aspekte der wirtschaftstheoretischen und industriepolitischen
Diskussion, einschließlich der Beziehungen zwischen den Herfindahl-Index und der
sozialen Wohlfahrt, freiem versus sozial optimalem Markteintritt, Konzentration und
Profitabilität, ruinösem Wettbewerb und natürlichem Monopol. Der analytische Ansatz
vereint Einfachheit mit Allgemeinheit.
1.
Introduction
There has been a rich and insightful debate in industrial organization about the welfare and
profitability effects of increasing the number of firms in an industry. While many complex
facets of this fundamental issue remain partly unsettled, the basic compromise at work is
well-known, though still a source of major controversy both among academics and antitrust
practitioners1 . On the one hand, conventional intuition — sometimes wrongly — holds that
increasing the number of firms reduces monopoly power and allows closer approximation
of the competitive ideal. On the other hand, increasing the number of firms may result in
reduced ability to take advantage of scale economies.
The relationship between concentration and profit rates has been one of the most active
research areas in empirical industrial organization (e.g. Mueller, 1986). While conventional
intuition holds that per-firm profit must decline with the number of firms, interaction between theoretical and empirical work has over the years uncovered the potential role of several
inter-related factors, including potential collusion, ease of entry, and merger policy. The importance of scale economies has been stressed early on to argue against antitrust-mandated
break-ups of large firms and for a laisser-faire policy (e.g. Demsetz, 1973).
The welfare implications of market structure have been prominent in the early beginning
of the field, Bain (1956). The unquestioned view then was that barriers to entry are responsible for the presence of imperfect competition, which in turn results in sizable welfare losses.
The belief that public policy must correct for this imperfection by removing barriers to entry and possibly subsidizing entry had dominated the profession and persisted until quite
recently. Perceptive work by von Weiszacker (1980), Perry (1984), Mankiw and Whinston
(1986) and Suzumura and Kiyono (1987) demonstrated that this view was fundamentally
ill-founded by showing that if firms’ conduct is not subject to regulation, then free entry
results in an excessive (endogenous) number of firms relative to a social optimum.
Once the existence of the trade-off between the benefits of competition and the costsavings of production scale is established, the main issue becomes largely empirical. Indeed,
the empirical literature on the subject is extensive, though somewhat outdated by now.
The current Microsoft case is a timely reminder of the importance of this issue, albeit with a relatively
new key dimension: Network effects.
1
1
Several studies in Goldschmid, Mann and Weston (1974) provide empirical evidence and a
general debate on this controversial issue. For instance, Sherer (1974) concludes on the basis
of a detailed empirical study of twelve manufacturing industries that the evidence provides
little support for the conjecture that concentration is the result of a realization of scale
economies. On the other hand, his findings are controversial (e.g. Brozen, 1973).
The present paper offers a thorough theoretical investigation of the effects on industry
profits and social welfare of exogenously increasing the number of firms in a Cournot industry
composed of identical firms. Given the importance that these issues have for a free market
society, the results presented here would be a pre-requisite to any modern empirical work or
formulation of public policy dealing with market structure and social welfare.
An extensive older literature addresses the related questions of the effects of the number
of firms on industry price and output levels, e.g. Ruffin (1973) and Seade (1980)2 . Yet, to
the best of our knowledge, no systematic theoretical analysis of industry profits and social
welfare has been conducted with an exogenous number of firms.
An important aspect of the paper, from a methodological viewpoint, is its reliance on the
new lattice-theoretic comparative statics approach. We build directly on Amir and Lambson
(2000) who use this same framework to analyse price and output effects. They derive two
main results. The first one is that industry price falls (increases) with the number of firms
if a firm’s residual inverse demand declines slower (faster) than its marginal cost, globally.3
This is the so-called property of quasi-competitiveness (quasi-anticompetitiveness). Strong
scale economies are required for demand to decline slower than marginal cost, and lead to
the counterintuitive outcome on price. The second result is that per-firm profit falls with
the number of firms in both cases.
A complementary methodological feature is our reliance on tight illustrative examples.
These serve a dual purpose. First, they confirm that the given sufficient conditions are, in
some sense, critical to the resulting conclusions. Second, they illustrate in a more accessible
manner the interaction between the various effects at work in the comparative statics at
hand. In particular, we present a Main Example (Section 3) which is a blueprint for the entire
See also Frank (1973), Okuguchi (1973) and Novshek (1980), among others.
This sharp intuition behind many of our main conclusion is only possible with the latice-theoretic approach.
In previous work, other method-imposed assumptions, such as concave profits and decreasing marginal
revenue, prevented such simple and clear-cut economic interpretations based only on critical assumptions.
2
3
2
paper, in that most effects of interest can be captured by varying the Example parameters.
This example can serve as a pedagogical tool to convey the main ideas of the paper in a
simple and intuitive framework to undergraduate students or policy practitioners.
For industry performance comparisons (Section 4), the quasi-competitive case has two
clear-cut welfare results that are independent of the nature of returns to scale: welfare
increases in the number of firms whenever per-firm output does, and in case of multiple
equilibria, the maximal output equilibrium is the social-welfare maximizing equilibrium. The
latter result vindicates the supremacy of consumer welfare over producer welfare. Otherwise,
this case gives rise to two subcases, depending on the returns to scale in production. In the
presence of economies of scale, industry profits are shown to globally decline with the number
of firms, while social welfare is generally not monotonic. More precisely, we argue that the
slightest amount of scale economies leads to welfare being decreasing at sufficiently high
number of firms. Inversely, under diseconomies of scale, social welfare is globally increasing
in the number of firms, while industry profits exhibit a tendency to initially increase in
the number of firms (treated as a real number), starting at monopoly level. (Whether this
tendency leads to duopoly, say, having higher total profit than monopoly depends on the
magnitude of the returns). As an important corollary of the two monotonicity results above,
under constant returns to scale, both conventional beliefs indeed hold: Industry profits fall
and social welfare increases with the number of firms.
For the quasi-anticompetitive case, monopoly always leads to the highest possible industry profits, with this being the only clear result on industry profits.4
On the other hand,
the welfare outcome is unambiguous: Social welfare always decreases in the number of firms.
In this case there are strong enough scale economies to overcome all other considerations.
Our conclusions provide a precise theoretical foundation for intuitive beliefs about the
need for a trade-off between the benefits of fostering increased competition and the ability
of firms to exploit scale economies. The conclusions are also fully congruent with some
classical results in partial equilibrium analysis. It is well-known (Ruffin, 1973) that, under
quasi-competitiveness, Cournot equilibria converge to perfectly competitive equilibria when
average cost is nondecreasing, and that this same convergence fails when average cost is
We provide an informal argument showing that, for intermediate market structures, the cost and revenue
effects move in conflicting directions, the latter in a counterintuitive way..
4
3
nonincreasing. Our results show that welfare is monotonically increasing in the former case
(thus converging to first-best welfare), but not in the latter. Also see Novshek (1980).
The results presented here may also be invoked to illuminate a number of important
theoretical and public policy debates: See Section 5 where some important corollaries of our
results are presented in various applications. The first of these concerns the relationship
between Cournot and perfectly competitive equilibria, as described above.
The second point deals with the comparison of endogenous concentration levels prevailing
under free and first and second-best socially optimal entry. Mankiw and Whinston (1986)
and Suzumura and Kiyono (1987) show that the old belief that free entry leads to socially
too few firms is invalid under general conditions. Our results provide simple but important
insights into this important issue in addition to clarifying the latter authors’ conclusions on
the comparison between free and first best entry.
The third point addresses the extensive use made by antitrust authorities of the HerfindahlHirschman index of concentration under the presumption that the Index is a good inverse
measure of social welfare. Recent theoretical work showed that with a constant number of
firms, any output transfers across firms that leave price unchanged must cause the Index
and social welfare to move in the same direction, Farrell and Shapiro (1990) and Salant and
Shaffer (1999). We complement this insight by the observation that with scale economies
and a varying number of firms, both the Index and welfare decrease with the number of
firms, when the latter is larger than a threshhold level, which may be one.
The fourth point considers the long-standing concentration/profitability debate, and relates our results to the inconclusive evidence uncovered over the years on this key issue. In
particular, in cases where free entry leads to a completely indeterminate number of firms
(the quasi-anticompetitive case) or to multiple equilibria with a given number of firms, any
meaningful correlation between industry characteristics and profits is unlikely.
The fifth point identifies the Cournot model in the quasi-anticompetitive case as an
appropriate framework for modelling the old concept of destructive competition. Indeed,
there is an excellent match between the theoretical predictions of the model in that case and
the stylised facts commonly associated with destructive competition (Sharkey, 1982).
The sixth point proposes to define natural (unregulated) monopoly as an industry where
4
the socially optimal number of firms is one, as opposed to the old definition of (regulated)
natural monopoly based on the inability to improve on costs by subdividing production,
a purely production-based criterion. This new definition clearly balances the market and
production sides, and is more appropriate in the absence of regulation and contestability.
All these applications emphasize the role of scale economies in engendering a trade-off
between the market effect and the production efficiency effect. They convey the sense that
our simple results form a pre-requisite for a thorough understanding of the issues presented.
2.
The Model
The fundamental questions under investigation here can be simply phrased as follows: How
do total equilibrium output (and hence industry price), per-firm profit, industry profit and
social welfare vary with the exogenously given number of firms in the industry? We consider
these fundamental questions in the framework of equilibrium comparisons (as in Milgrom
and Roberts, 1994), the exogenous parameter being the integer number of firms.
We begin with the basic notation. Let P : R+ → R+ be the inverse demand function,
C : R+ → R+ the (common) cost function, A : R+ → R+ the average cost function, and n
the number of firms in the industry. Let x denote the output of the firm under consideration,
y the total output for the other (n − 1) firms, and z the cumulative industry output, i.e.,
z = x+y. At equilibrium, these quantities will be indexed by the underlying number of firms
n. We explicitly deal with the (possible) nonuniqueness of Cournot equilibria by considering
extremal equilibria. Denote the maximal and minimal points of any equilibrium set by an
upper and a lower bar, respectively. Thus, for instance, z n and z n are the highest and lowest
total equilibrium outputs, with corresponding equilibrium prices pn and pn , respectively.
Performing comparative statics on equilibrium sets will consist of predicting the direction of
change of these extremal elements as the exogenous parameter varies.
The profit function of the firm under consideration is
Π (x, y) = xP (x + y) − C (x)
(1)
Alternatively, one may think of the firm as choosing total output z = x + y, given the
5
other firms’ cumulative output y, in which case its profit is given by
e (z, y) = Π (z − y, y) = (z − y) P (z) − C (z − y)
Π
e with respect to z and y, i.e.,
Let 4 (z, y) denote the cross-partial derivative of Π
0
00
4 (z, y) = −P (z) + C (z − y)
(2)
(3)
∧
e and 4 are defined on the set ϕ =
Note that both Π
{(z, y) : y ≥ 0, z ≥ y}.
The following Standard Assumptions are in effect throughout the paper:
0
(A1) P (·) is continuously differentiable and P (·) < 0.
0
(A2) C (·) is twice continuously differentiable and C (·) > 0.
(A3) There exists x
b > 0 such that P (x) < A(x) for all x > x
b.
Although convenient, the smoothness assumptions are by no means necessary for our main
results. (A3) simply guarantees that that a firm’s reaction curve eventually coincides with
the horizontal axis, so that a firm’s effective outputs, and thus all Cournot equilibrium
outputs, are bounded by some constant, say K , for all n.
The qualitative nature of most of our results hinges entirely on the global sign of ∆, so
that we will distinguish two main cases: ∆ > 0 and ∆ < 0. When ∆ > 0 globally, there will
be two subcases of interest depending on the returns to scale, or the slope of the average
cost curve. This division is already apparent in the upcoming example, which may serve as
a blueprint for the entire paper.
3.
Returns to Scale, Concentration and Industry Performance:
The Main Example
We now consider a simple example that provides an excellent and thorough overview of
most of the results derived in this paper.5
As a parameter capturing the returns to scale
is varied, the example can fit the two major cases of analysis of the general model: ∆ < 0
and ∆ > 0. In the latter case, the example can also capture the two subcases of interest:
economies or diseconomies of scale6 . In addition, this example will also be invoked later on
This example would be very appropriate for the purpose of presenting in a very elementary framework the
essentials of the analysis to undergraduate students or economic practitioners.
6 This specification of the demand and cost functions with the further restriction that d < 0 has already
been considered with different motivations in Cox and Walker (1998) and d’Aspremont, Gerard-Varet and
Dos-Santos-Ferreira (2000). Neither study deals with the main issues under consideration here, namely
industry profits and social welfare.
5
6
to gain further insight into the tightness of the conditions behind our general results.
Let the inverse demand be linear and the cost function be quadratic, i.e.,
P (z) = a − bz
and
C(x) = cx + dx2
(4)
with the assumption throughout that a > c > 0 , b > 0 , b + d > 0 .
Since A(x) = c + dx, returns to scale are increasing (decreasing) if d < (>)0. Thus d is
our returns to scale parameter, key to many results below. Furthermore (cf. (3))
∆ = b + 2d R 0 if d R −b/2
The reaction function is always linear (when strictly positive) and given by r(y) =
(5)
a−c−by
.
2(b+d)
Thus for any number of firms n there is always a unique symmetric Cournot equilibrium.
Omitting some lengthy calculations (including solving for the symmetric Cournot equilibrium
via r[(n − 1)xn ] = xn ), this equilibrium has per-firm output and profit, and social welfare
given respectively by
xn =
a−c
(b + d)(a − c)2
n[b(n + 2) + 2d)](a − c)2
, πn =
,
W
=
n
b(n + 1) + 2d
[b(n + 1) + 2d]2
2[b(n + 1) + 2d]2
(6)
Furthermore, if d > −b/2 (or ∆ > 0), the slope of the reaction curve in larger than −1 and
the symmetric equilibrium is the unique equilibrium. It is also globally stable in the sense
that best-reply Cournot dynamics converges to this equilibrium, from any initial outputs.
On the other hand, if d < −b/2 (or ∆ < 0), the slope of the reaction curve is smaller
than −1, so that r(y) decreases steeply and is equal to 0 when y ≥ (a − c)/b . Consequently,
there are several other Cournot equilibria, and they can be characterized as follows. With
n being the total number of firms in the industry, if any m firms (with m < n) produce the
output xm each (given in (6)), and the remaining n − m firms produce nothing, the resulting
output configuration is clearly a Cournot equilibrium.7 For the n-firm oligopoly, the unique
symmetric equilibrium (with all firms active) is unstable in the sense that best-reply Cournot
dynamics diverge away from it (Seade, 1980).
It can be verified that
7 To see this, observe that r(mxm ) = 0, since mxm ≥ (a−c)/b, as can be easily checked. In particular, if any
one firm produces the optimal monopoly output x1 = (a − c)/(b + 2d), and all the others produce nothing,
we have a Cournot equilibrium. Again, this follows from r(x1 ) = 0 since x1 ≥ (a − c)/b, and x1 = r(0)
clearly. Given the linearity of the reaction curve here, this is easy to see graphically.
7
(i) per-firm output xn is always decreasing in n (cf. Proposition 1b).
(ii) industry output zn = nxn is increasing in n if d > −b/2 (or ∆ > 0) and decreasing in n
if d < −b/2 (or ∆ < 0) (cf. Propositions 1a and 2a).8
(iii) per-firm profit π n is always decreasing in n (cf. Propositions 1c and 2c).
It remains to analyse the effects of n on industry profits and social welfare. It is convenient
here to treat the number of firms as a real variable. For industry profits, we have (with the
computational details left out)
∂(nπ n )
T 0 if and only if n S 1 + 2d/b.
∂n
(7)
Here, there are two separate cases of interest:
(i) d < 0 : Then industry profits always decrease with the number of firms, with in particular
monopoly having the largest industry profit.
(ii) d > 0 : Then n
e = 1 + 2d/b maximizes industry profits, which thus increase in n when
n<n
e and decrease in n when n > n
e, starting from any given n. Observe that if d/b < 1/2,
then 1 < n
e < 2. Hence, in particular, if n
e = 1 (i.e. monopoly is the market structure that
maximizes total profits), industry profits would be globally decreasing in n. But if d is large
enough, i.e., if there are sufficiently high returns to scale, industry profits will be rising in n
initially, all the way to n
e which may be a large number of firms, but industry profits always
eventually decrease in n (i.e., for n > n
e.)
For social welfare, one can easily verify that
∂Wn
T 0 if and only if nbd T −(b + d)(b + 2d)
∂n
(8)
Again, there are two separate cases of interest:
(a) ∆ < 0 iff d < −b/2: Welfare always decreases with n (cf. Proposition 10.)
(b) ∆ > 0 iff d > −b/2: Here, there are two different subcases of interest.
(i) d > 0 : welfare always increases in n (cf. Proposition 6.)
(ii) if −b/2 < d < 0, then welfare is maximized at n∗ = −(b + d)(b + 2d)/bd , increases in
n for n < n∗ and decreases in n for n > n∗ (cf. Proposition 8). Observe that this statement
The intuition behind the counter-intuitive case ∆ < 0 is that with more competition, each firm lowers
output drastically since r0 (y) < −1, thereby moving up its steeply declining average cost curve. The resulting
efficiency loss is large enough to overcome the downward pressure on price engendered by the increase in
competition. The increase in average cost is passed on to consumers via a higher price.
8
8
is true no matter how close d is to 0 (from below)! In other words, the slightest presence of
(uniform) scale economies causes welfare to be eventually declining in n (i.e., for sufficiently
large values of n). The parameters of this example can be chosen to make n∗ equal any
desired value from 1 on, while satisfying all the underlying constraints here.
The economic intuition can now be stated concisely and precisely since the main results
hinge mainly on the sign of ∆ = −P 0 + C 00 , and sometimes also on the returns to scale.
For industry price, there are two effects at work, a market or competition effect captured
by the term −P 0 , and a production or scale effect captured by C 00 . The market effect always
pushes in the intuitive direction that price should fall with the number of firms. The scale
effect goes in the same direction if and only if costs are convex. When costs are concave, the
overall outcome on price is determined by the relative strength of the two effects. Per-firm
profit always behave in the intuitive way.
For industry profits, the market effect pushes in the intuitive direction if and only if
industry price is well-behaved ( ∆ > 0). The production effect works in the intuitive direction
if scale economies are present. When the two effects are antagonistic, the outcome depends
on the relative strengths again.
Viewed as the sum of consumer and producer surpluses, social welfare can be discussed
on the basis of the previous assessments. Thus, with ∆ > 0 and diseconomies of scale,
consumer surpluse increases with n, overcoming a possible decrease in producer surplus (the
latter effect being ambiguous). With ∆ < 0, strong scale economies are necessarily present,
and both surpluses decrease wth n. Finally, with ∆ > 0 and economies of scale, consumer
surplus moves up and producer surplus down, with an ambiguous net effect.
In conclusion, this example provides a microcosm for the entire paper. In the remainder,
we simply present a generalization of the insights illustrated so far, preserving another key
role for this example in testing the tightness of the sufficient conditions given for our various
results. The less theoretically-minded reader may jump to the last section for a summary of
the economic implications of the results, as conveyed by the Main Example.
9
4.
A General Cournot Analysis of Industry Performance
This section contains the general analysis of the interplay between market structure or
concentration and scale economies in determining industry performance as reflected in price,
outputs, industry profits and social welfare. This amounts to comparing Cournot equilibria
along these characteristics as the number of firms varies. In an attempt to obtain the broadest
possible understanding of the issues involved, we provide a series of minimally sufficient
conditions for the desired conclusions, combined with tight complementary examples to
shed further light on the relationship between assumptions and conclusions. The proofs
combine analytical simplicity with generality. Nonetheless, we supplement the presentation
with some heuristic arguments whenever it is felt they may provide additional insight.
Methodologically, we make crucial use of the lattice-theoretic comparative statics approach.9 This allows for very general conclusions relying only on critically needed assumptions, thereby leading to clean and tight economic interpretations of the conclusions, as well
as analytical rigor. In the present context, the usual arguments in favor of this approach
become even more pertinent, as the parameter of interest, the number of firms, is an integer:
See Appendix for a graphical illustration of many important details on this issue.10
While the condition ∆ > 0 is familiar in Cournot theory at least since Hahn (1962), it
has typically been used in conjunction with many other assumptions, such as some form of
concavity of each firm’s profit in own output, decreasing marginal revenue, etc...The latter
assumptions interfere with a good intuitive understanding of the economic forces at work,
as they are also made for the case ∆ < 0. As shown below, there is a very natural division
of the results here, and it depends only on the global sign of ∆. The latter has a very
simple and appealing interpretation: ∆ > 0 (∆ < 0) means that price, or residual inverse
0
demand, decreases (increases) faster than marginal cost. Since P < 0, it is clear that the
In particular, we invoke the general results of Milgrom and Roberts (1990, 1994), in addition to Topkis
(1978). More specific to Cournot oligopoly, we build on the results of Amir (1996) and Amir and Lambson
(2000). See also McManus (1962) and Roberts and Sonnenschein (1976) for early antecedents.
10 The traditional method based on the Implicit Function Theorem and the signing of derivatives can provide
insight for special cases, but is ill-suited for the requisite analysis at hand. The main reason is that it rests
on assumptions (such as concavity and equilibrium uniqueness) that need not be satisfied in our general
setting. Furthermore, as these same assumptions are needed in all the otherwise mutually exclusive cases
of analysis, this method prevents a tight intuitive understanding of the economic forces behind each result.
Finally, in the context at hand, as the parameter of interest is an integer, traditional methods have further
drawbacks: See de Meza (1985), Amir and Lambson (2000) and Milgrom and Roberts (1994).
9
10
convexity of C implies 4 > 0 on ϕ. Likewise, strong concavity of C is required for ∆ < 0.
Examples provided below illustrate that ∆ > 0 can hold globally even when the cost function
is everywhere concave, an important subcase of analysis in this paper.
4.1
Equilibrium Price and Outputs
The results of this subsection have been proved in Amir and Lambson (2000). They are
stated here without proof, and used in the sequel in looking at industry profit and social
welfare. For a more detailed presentation, the reader is referred to the above paper. In the
Appendix, a graphical illustration of the need for the new comparative statics is presented,
with the conclusion that only extremal equilibria can be unambiguously compared as the
number of firms varies.11
For any variable of interest, the maximal (minimal) value will
always be denoted by an upper (lower) bar.
Proposition 1 Let 4 (z, y) > 0 on ϕ. For each n, there exists a symmetric equilibrium and
no asymmetric equilibria. Let xn be an extremal Cournot equilibrium output.
(a)Industry output zn is nondecreasing in n, and hence price pn is nonincreasing in n.
(b) xn is nonincreasing [nondecreasing] in n if log P is concave [convex and C(·) ≡ 0].
(c)The corresponding equilibrium profit π n is nonincreasing in n.
Thus the Cournot model is quasi-competitive here (Part (a)). The fundamentally needed
e on ϕ, which is equivalent to ∂ 2 Π/∂z∂y
e
assumption is the supermodularity of Π
= ∆ > 0.
This implies that the line segment joining any two points on the graph of the reaction
correspondence r of a firm must have a slope ≥ −1, which means that, in response to an
increase in rivals’ output, a firm can never contract its output by more than this increase.
In particular, this precludes downward jumps for r (while allowing for upward jumps).12
The other case is characterized by the assumption 4 (z, y) < 0, implying that r has all its
slopes bounded above by −1: as the joint output of the rivals is increased, a firm optimally
reacts by contracting its output so much that the resulting total output decreases. Hence,
r is (strongly) decreasing, so that the results of Novshek (1985) and Amir (1996) guarantee
the existence of a Cournot equilibrium.13 However, existence of a symmetric equilibrium is
11 For
unstable equilibria (in the sense of best-reply Cournot dynamics), the price comparative statics is
counter-intuitive (as seen in the Appendix), and this will carry through to other results.
12 This property was noticed and exploited in classic papers by McManus (1962) and Roberts and Sonnenschein (1976) to establish the existence of Cournot equilibrium in the case of symmetric firms with convex
cost functions. See Amir (1996) and Amir and Lambson (2000) for further generalization.
13
11
not guaranteed for all n, unless we assume in addition quasi-concavity of Π(·, y): see below.
To separate the issues of existence and of comparing equilibria, we state the result without
this assumption. Let N be the subset of n’s for which a symmetric equilibrium exists (i.e.
those n’s for which r intersects the line y/(n − 1).)
Proposition 2 Let 4 (z, y) < 0 on ϕ and x1 ≤ K . Then for any given n ≥ 2,
(a) If n ∈ N, then the symmetric equilibrium is unique and satisfies: xn , zn and πn are
nonincreasing in n. Hence pn is nondecreasing in n.
(b) For any m ∈ N with 1 ≤ m < n, the following is an equilibrium for the n-firm oligopoly:
Each of any m firms produces xm while the remaining (n − m) firms produce nothing. All
these Cournot equilibria are invariant in n, in that all entering firms would produce zero.
(c) No other Cournot equilibrium (than those of Parts (a) and (b)) can exist.
The minimal assumption to get existence of a symmetric equilibrium for all n is that Π (x, y)
is strictly quasi-concave in x for every y ∈ [0, y], where14 y = P −1 [C 0 (0)]. Otherwise, the
best-response may have a downward jump where it skips over the y/(n − 1) lines for some
n’s, thus implying the absence of a symmetric equilibrium for those n’s.
4.2
Industry Profits and Social Welfare
Here, the effects of an exogenous change in the number of firms on total profits and social
welfare are investigated. Most of the results below are stated both in a local and in a global
sense, the latter being a direct consequence of the former. Elementary (heuristic) proofs for
the global statement are given in the text, while rigorous proofs for the local statement are
given in Appendix. We emphasize that the latter, while not standard, nonetheless combine
generality with simplicity! For the same reasons as before, we continue to focus on the
two extremal equilibria for all our results, and to separate the analysis of our comparativeequilibria results into two cases, according to the global sign of ∆.
4.3
The Case ∆ > 0
Recall that ∆ > 0 globally is consistent with both globally increasing and decreasing returns
to scale. We begin with the effects of concentration on industry profits. The local statement
However, there is no guarantee a priori that this includes more than the monopoly equilibrium (which
always exists).
14 By l’Hospital’s rule, P (y) = C 0 (0) = A(0). So y is the cumulative output of the other firms that equates
price and average cost when the firm produces nothing. Or y is the smallest y for which r(y) = 0.
12
(a) below is more general than the global statement (b). The former requires a rigorous
proof, given in Appendix (A1), while the latter is heuristically proved here with a standard
approach. This separation of proofs applies to most of our main results below.
Proposition 3 Let ∆ > 0 on ϕ. For the extremal equilibria,
(a)Industry profit nπ n is globally nonincreasing in n if A(·) is nonincreasing.
(b)nπ n ≥ (n + 1)π n+1 for any given n if A( n+1
xn+1 ) ≤ A(xn+1 ).
n
Proof. We present here a simple heuristic proof of (a). Assuming n is a real and nπ n =
zn [P (zn )−A(xn )] is differentiable in n, one can get
d(nπ n )
dn
=
dzn
n
[P (zn )−A(xn )]+zn [P 0 (zn) dz
−
dn
dn
n
A0 (xn ) dx
]. Using the first-order condition for a symmetric Cournot equilibrium,
dn
P (zn ) − A(xn ) + xn [P 0 (zn ) − A0 (xn )] = 0
(9)
and simplifying, one arrives at
dzn
d(nπ n )
= xn A0 (xn ) + (n − 1)P 0 (zn )
dn
dn
Now,
d(nπ n )
dn
≤ 0 follows from A0 (xn) ≤ 0 and
dzn
dn
(10)
≥ 0 since ∆ > 0 (Proposition 1a).
Since per-firm profit π n always falls with n, Proposition 3 asks whether π n falls fast
enough to have nπ n ≥ (n + 1)π n+1 . In interpreting the proposition, it is convenient to
separate the overall effect of an increase in the number of firms on industry profits into two
distinct parts, as suggested by the above proof. The market or total revenue effect, which
may be isolated by setting A0 = 0, always pushes in the intuitive direction that industry
profits must fall. On the other hand, the production or efficiency effect goes in the same
direction if and only if scale economies are present.15
The proof of the result also makes it clear that the conclusion follows when both the
market and the production efficiency effects push in the same direction, which suggests the
condition on average cost is sufficient but not necessary. It is natural then to ask how critical
this assumption is for this conclusion. Treating the number of firms as a real number, the
following argument provides a simple but interesting insight: Monopoly is never the profit
maximizing market structure under increasing average cost.
15 By
contrast, an n-firm cartel always has higher optimal profit than the total n-firm oligopoly profit, since
the cartel, with access to n plants, always has the option of producing nxn at a cost at most equal to the
total cost of the n-firm oligopoly. There is thus an obvious difference between a monopoly (with access to
one plant) and a cartel composed of n identical firms.
13
Proposition 4 if A0 (·) > 0, industry profit increases from monopoly level as the number of
firms is increased slightly beyond n = 1.
Proof. Setting n = 1 in (10), we have
h
d(nπ n )
dn
i
n=1
= x1 A0 (x1 ) > 0.
Observe that this need not mean that duopoly has higher profit than monopoly, as
industry profit may peak between n = 1 and n = 2, with either π1 or 2π 2 as the highest
value. This point is illustrated in the Main Example where, for d < b/2, industry profits
may well be globally decreasing in the number of firms, and are certainly decreasing in n for
n ≥ 2 (see (7)). Thus, Proposition 4 relies crucially on the number of firms being a real.
Nonetheless, the point made here is important as it shows that the slightest amount
of increasing returns pushes toward industry profits that are increasing in the number of
firms. Whether this effect actually succeeds in preventing industry profit from being globally
decreasing in the integer number of firms depends on the strength of the increasing returns,
as suggested by the Main Example. Indeed, from (7), a sufficient condition for industry
profit not to be globally decreasing in n is d > b/2.
We now turn to the welfare analysis. It can easily be shown that xn is the Paretodominant equilibrium for the firms (i.e., leads to the largest producer surplus) while xn
is the Pareto-preferred equilibrium for the consumers (i.e. leads to the largest consumer
surplus). It is then of interest to know whether the Cournot equilibria are ranked according
Rz
to the Marshallian measure of social welfare, defined as 0 P (t)dt−nC(z/n). In other words,
is one of the two surpluses always dominant? The next result (proof in Appendix A1) settles
this question in favor of consumer surplus.
Proposition 5 Let ∆ > 0, and xn and x0n denote two distinct equilibrium per-firm outputs
with corresponding social welfare levels Wn and Wn0 . If xn ≤ x0n , then Wn ≥ Wn0 . Hence, xn
is the social welfare maximizer among all equilibrium per-firm outputs.
As Proposition 1 shows, the case ∆ > 0 is consistent with both xn increasing and xn
decreasing. The implications of these two possibilities on social welfare are quite different,
as reflected in the next result. Also, if the demand function does not satisfy either condition
(log-concavity or log-convexity) from Proposition 1(b), then xn will generally not be monotonic in n. Thus, in the following result, the local statement (b) of the welfare result is more
general than the global statement (a).
14
Proposition 6 Let ∆ > 0 on ϕ. For any n, at an extremal equilibrium,
(a) Social welfare is nondecreasing in n if (i) A(·) is nondecreasing and xn ≥ xn+1 , or
(ii) xn ≤ xn+1 .
(b) Wn+1 ≥ Wn for a given n if either one of the following holds: (i) A(xn+1 ) ≤ A(xn),
or (ii) xn ≤ xn+1 .
Proof. We prove Part (a) in a simple heuristic way, and leave the proof of Part (b) to
Appendix. Assuming differentiability of Wn and xn with respect to n, and differentiating
Rz
n
n
n
n
through Wn = 0 n P (t)dt − zn A(xn ), one gets dW
= P (zn ) dz
− dz
A(xn ) − zn A0 (xn ) dx
, or
dn
dn
dn
dn
dWn
dxn
= π n + n[P (zn ) − C 0 (xn )
]
dn
dn
(11)
Using the first-order condition for a Cournot equilibrium (9), and simplifying yields
dWn
dzn 0
= xn [A0 (xn )xn −
P (zn )]
dn
dn
For (a), the conclusion follows from (11) since P (zn ) ≥ C 0 (xn ) and
conclusion follows from (12) since A0 (xn ) ≥ 0 and
dzn
dn
(12)
dxn
dn
≥ 0. For (b), the
≥ 0 (from ∆ > 0).
Since price falls with the number of firms here, consumer surplus always increases. However, producer surplus may a priori move either way. So the proposition identifies two
sufficient conditions (diseconomies of scale and decreasing per-firm output, or increasing
per-firm output) implying that total profit will never decrease enough to overcome the increase in consumer welfare and result in lower social welfare. An alternative way to think of
this result is as follows. Due to the increase in industry output, the sum of consumer surplus
and industry revenue (i.e. total benefit or the total area under the inverse demand up to the
equilibrium output) always increases with the number of firms. On the other hand, industry
costs may go either way. In this perspective, Proposition 6 identifies conditions ensuring
that industry costs will never increase enough to cause social welfare to overall decrease, in
spite of the increase in total benefit.
Propositions 3 and 6, taken together, imply that conventional wisdom fully prevails for
the case of constant returns to scale, which is widely invoked in industrial organization.
Corollary 7 With linear cost, C(x) = cx, at an extremal equilibrium, nπ n is nonincreasing
in n and social welfare Wn is nondecreasing in n, for all n.
Proof. This follows directly from Propositions 3 and 6, as average cost is constant.
15
In the presence of scale economies, it is of interest to shed further light on the welfaremaximizing number of firms n∗ . Treating the number of firms as a real variable, we can
provide a heuristic derivation16 and an interesting interpretation of n∗ (cf. (8)).
Proposition 8 Assume ∆ > 0 and global scale economies prevail (i.e. A0 (·) ≤ −ε < 0).
Then a necessary condition for a finite welfare-maximizing number of firms n∗ is that
d[P (zn )]
= C 0 (xn) − A(xn )
dn
(13)
Proof. It is convenient here to rewrite (12) as
dWn
1 dzn 0
= x2n [A0 (xn ) −
P (zn )].
dn
xn dn
(14)
Differentiating the first-order condition P (zn) + xn P 0 (zn ) = C 0 (xn ) w.r.to n and symplifying,
one can evaluate
1 dzn
P 0 (zn ) − C 00 (xn )
=
−→ 0 as n → ∞.
xn dn
(n + 1)P 0 (zn ) + zn P 00 (zn ) − C 00 (xn )
n
Hence, there is some N large enough such that n ≥ N =⇒ − x1n dz
P 0 (zn ) < ε ≤ −A0 (xn ),
dn
and xn > 0, so that
dWn
dn
dWn
dn
< 0 from (14). This says Wn has a maximum for n ≤ N . Setting
n
= 0 in (14) yields P 0 (zn ) dz
= xn A0 (xn ). (13) follows then from the obvious relations
dn
C 0 (xn ) − A(xn ) = xn A0 (xn ) and
d[P (zn )]
dn
n
= P 0 (zn ) dz
.
dn
Thus, the socially optimal number of firms has a simple interpretation: The last ”marginal
firm” taken in lowers industry price by exactly the difference between average and marginal
cost at the equilibrium per-firm output. This equalizes marginal social benefit (the sum
of consumer surplus and firms’ revenues) with marginal social cost (the production costs).
While the fact that the socially optimal number of firms is typically finite in the presence of
fixed costs is well-known, Proposition 8 is nonetheless somewhat surprising as it relies only
on the slightest level of scale economies.17
The next example shows that if A0 ≤ 0 but A0 (0) = 0, Wn may be globally increasing in
n, so that n∗ = ∞. Hence, the assumption A0 (0) > 0 in Proposition 8 is crucially needed:18
16 Again,
this derivation is based on the implicit function theorem, which cannot be justified for the present
use, in view of the multiplicity of equilibria, among other problems (see Appendix A1.)
17 With diseconomies of scale (or A0 (·) > 0), (13) holds with a ≤ sign instead of the = sign. In the limit as
n → ∞, both sides of (13) are zero, as should be the case for perfect competition (recall that marginal and
average costs intersect at the latter’s minimum): See Ruffin (1971).
18 The Main Example also shows that any (uniform) level of scale economies, i.e. the smallest (in absolute
value) d < 0, the conclusion that social welfare globally increases with the number of firms would fail as
shown by (8): See Point (b)(ii) just below (8).
16
Example 2. Let P (z) = 2 − z and C(x) = x − .1x3 /3, for x ≤
√
30.
Thus A(x) = 1 − .1x2 /3 and A0 (0) = 0. There is a unique Cournot equilibrium with
p
xn = 5[n + 1 − (n + 1)2 − .4]. It can be numerically verified that Wn is increasing in n.19
4.4
The case ∆ < 0
Strong economies of scale are necessary for ∆ to be globally negative. One feature that is
known to give rise to economies of scale is the presence of (avoidable) fixed-costs. Without
these, one needs a strongly concave cost function for ∆ < 0 to be possible.
As existence of a symmetric Cournot equilibrium with all firms active is not guaranteed
for all n, there are two meaningful ways to proceed. One is to assume quasi-concavity of each
profit function in own output and restore existence. The other is to view the comparative
statics results as holding for those n’s for which the symmetric equilibrium exists.
In view of the multiplicity of Cournot equilibria described in Proposition 2, free entry
would give rise to every possible integer number of firms being active, with all firms producing
equal outputs.20
Furthermore, the equilibrium with m active firms is also a Stackelberg
equilibrium of a two-stage game where the m firms act as first-movers and the rest of the
firms as second movers, m ≤ n − 1: See Amir and Lambson (2000) and Robson (1990).
The only general result on industry profit we can offer here vindicates the conventional
wisdom only about monopoly.
Proposition 9 Let ∆ < 0 on ϕ. Industry profit is highest under a monopoly than under
any other market structure, i.e., π 1 ≥ nπ n , ∀n.
Proof. Since the cost function is concave (hence subadditive), a single firm has the option
of producing the n-firm total Cournot output zn = nxn at a cost lower than that of the
n-firm oligopoly (i.e., C(nxn ) ≤ nC(xn )) for any n. The conclusion then follows.
While no counterexample could be found to establish that nπ n is not always decreasing
in n, the following argument suggests the conjecture might be false. Total cost is easily seen
to increase in n here, but the revenue part ”moves in the counterintuitive direction”.21
q
q
q
2
2
2
2
Wn = 5x(x + 1 − (x + 1) − .4) − 12.5x (x + 1 − (x + 1) − .4) + 4. 166 7x(x + 1 − (x + 1)2 − .4)3 .
20 This refers to the subgame-perfect equilibria of a two-game of entry where (infinitely many) firms simultaneously decide whether to enter or not at no cost in the first stage, and the entrants then engage in Cournot
competition in the second stage upon observing the number of entrants. Lopez-Cunat (1999) analyses the
differences between this entry process and the one-stage version used by Novshek (1980) among many others.
19
21
17
The welfare comparative statics is unambiguous here, due to the strong scale economies:
With more firms, output per firm is strongly reduced, resulting in a drastic increase in
average cost. This efficieny loss overcomes any other countervailing considerations.
Proposition 10 Let ∆ < 0 on ϕ. Then at the unique symmetric equilibrium, social welfare
Wn is nonincreasing in n, for all n.
Proof. The conclusion follows from (12) since A0 (xn ) ≤ 0 and
dzn
dn
≤ 0 (from ∆ < 0).
Since price increases with the number of firms here (Proposition 2a), consumer surplus
decreases. Also, as average cost and equilibrium per-firm output both decline rapidly, equilibrium total production costs increase rapidly with the number of firms here. Hence, even if
total profits go up, the increase will never be sufficient (recall also that per-firm profit goes
down) to overcome the fall in consumer surplus.
Corollary 11 Let ∆ < 0 on ϕ. Then monopoly leads both to the highest producer surplus
and to the highest consumer surplus levels.
Proof. The two statements follow respectively from Propositions 9 and 2a.
In view of the counterintuitive nature of many of the results in the case ∆ < 0, it is natural
to ask whether these results have any predictive value in describing imperfect competition in
some real-world markets.22
Experimental evidence suggests that unique (stable) Cournot
equilibria are good predictors of actual behavior (Holt, 1986). By contrast, Cox and Walker
(1998) report that in a symmetric Cournot game with three equilibria, a symmetric unstable
one and two boundary or monopoly equilibria (cf. Proposition 2), laboratory behavior
reflected no regular patterns of play that would support any of the three equilibria. Rather,
play seemed to proceed along irregular cycles around the three equilibria, meaning that the
players continuously exhibited large swings in their output levels, conveying a clear sense of
unstable behavior. On the other hand, none of the Nash equilibrium refinements for oneshot games (such as perfection, properness, strategic stability, etc..., see e.g. Fudenberg and
Indeed, if it were possible to have nxn ≥ (n + 1)xn+1 while C ≡ 0 (which we know ³
is impossible since
´
nxn
nxn
C ≡ 0 clearly implies ∆ > 0), we would have Πn+1 = xn+1 P [(n + 1)xn+1 ] ≥ n+1
P n+1
+ nxn+1 ≥
´
³
nxn
nxn
nxn
nxn
n
= n+1
P
+
n
P (nxn ) = nΠ
n+1
n+1
n+1
n+1 , where the first inequality is from the Cournot equilibrium
property and the second from the facts that P is decreasing and nxn ≥ (n + 1)xn+1 . It would then follow
that (n + 1)Πn+1 ≥ nΠn : Industry profit would be increasing in the number of firms!
22 Further discussion about this case is provided in Section 5 where the characteristics of the Cournot equilibria here are identified with the old concept of destructive competition, among other applications.
18
Tirole, 1991) can discard Cournot-unstable equilibria, although some evolutionary learning
processes can. Finally, regardless of stability properties, symmetric Cournot equilibria can
always be regarded as focal (Schelling, 1960.)
4.5
Hybrid Cases
In view of the level of generality of our conclusions, the fact that the entire analysis rests
essentially on one easily checked condition on the global sign of ∆ is a remarkable feature. On
the other hand, there are many demand-cost combinations of interest for which ∆ changes
signs on its domain: Hybrid cases. For these, Cournot equilibria will generally not behave in
the globally monotonic ways we uncovered here. The issue of existence also needs separate
attention then. De Meza (1985) provides an interesting hybrid counterexample highlighting
the differences between local and global comparative statics and showing that treating n
as a real number can lead to misleading results. Nonetheless, in spite of the nonmonotonic
behavior of some of the variables of interest, some of the insights we developed can still be
useful here. We offer an example.
Example 3. Let P (z) =
1
z+1
and C(x) = 12 log(x + 1).
It is easily checked that 4 (z, y) changes signs on ϕ (so that our results do not apply here),
x
1
and that Π(x, y) =
− Log (x + 1) is quasi-concave in x, for fixed y. The reaction
p x+y+1 2
1
curve is r(y) = 1 − y 2 , and the (unique) Cournot equilibrium is given by xn = √n2 −2n+2
.
Simple calculations show that while xn and nπ n are decreasing in n, for all n, zn and thus
pn are not monotonic in n (details are left out).
n
1
Social welfare is given by Wn = log( √n2 −2n+2
+ 1) − n2 log( √n2 −2n+2
+ 1). In particular,
√
¡√
¢
¡
¢
W1 = 12 ln 2 ' . 34657, and W2 = ln 2 + 1 − ln 12 2 + 1 ' . 34657, so that W2 ' W1 .
Thus23 , n∗ = {1, 2}: A social planner is indifferent (at least up to 5 decimals) between
monopoly and duopoly as the optimal choice!
5.
On some Theoretical and Policy Implications
The results presented here lie at the heart of the modern theory of industrial organization
and can, to some extent, illuminate a number of past as well as present theoretical issues
23 In fact, viewing n as a real variable, it can be shown that Wn is single-peaked in n and achieves its
maximum at n ' 1.36.
19
and public policy debates. In particular, we relate our findings to the relationship between
Cournot outcomes and perfect competition, the regulation of entry, the welfare content of
the Herfindahl index, natural monopoly and destructive competition. Surprisingly, the latter
two notions have not really been linked with Cournot theory in the past. We attempt to fill
this gap below. In some cases, we also present some new results here.
5.1
Relationship to Perfect Competition
Ruffin (1971) showed that if the number of firms is increased with fixed demand24 , Cournot
equilibria converge to the perfectly competitive equilibrium under global diseconomies of
scale, but not under global economies of scale. Our conclusions shed some light on this
result by indicating that (i) in the former case, equilibrium welfare converges monotonically
to first-best welfare, and (ii) in the latter case, although industry profits and per-firm output
both monotonically converge to zero, welfare does not increase to first-best welfare, due to
firms producing at increasing (and in the limit, maximal) average cost. Here first-best welfare
would involve one firm producing the entire output and pricing at marginal cost.25
5.2
Free versus Socially Optimal Entry
Perry (1984), Mankiw and Whinston (1986) and Suzumura and Kiyono (1987) put to rest a
long-standing belief that free-entry leads to insufficient (unregulated) competition by showing
that the opposite conclusion holds under general assumptions, using a two-stage game where
firms decide upon entry in the first period, and then compete a la Cournot in the second
period. The comparison benchmark is second-best social optimum in that the regulator
controls firms’ entry decisions but not their market conduct. Suzumura and Kiyono (1987)
differs from the other two papers in that costless entry is assumed, and two extra results
are established: (i) with convex costs, free entry is excessive relative to the first-best level
(where the regulator also controls pricing or market conduct, see Footnote 25), and (ii) the
first and second-best levels of entry are generally not comparable.
Our welfare results indicate that some of Suzumura and Kiyono’s excess entry findings
24 See Novshek (1980) for the other approach, where demand is replicated.
R nx
25 The planner’s objective is then max{
0 P (t)dt − nC(x) : n ≥ 1, x ≥ 0}.
0
The first-order conditions are
P (nx) = C (x) and xP (nx) = C(x). These imply that marginal and average cost are equal, as is well-known
for a first-best solution. In particular, the first-best number of firms is then 1 if A0 < 0 and ∞ if A0 > 0.
20
are not really instructive, and potentially misleading. Indeed, with diseconomies of scale,
Cournot equilibrium welfare increases monotonically with the number of firms to the firstbest level, so that both the first and the second-best socially optimal numbers of firms are
infinite (first-best welfare is, by definition, higher than second-best welfare. See Footnote
25.) The free-entry number of firms is also infinite 26 , so that all three entry levels are
actually equal in a trivial way.
With (strict) economies of scale27 , the second-best socially optimal number of firms is
finite, and equal to n∗ (which is 1 if ∆ < 0), the first-best is always 1 (see footnote 25),
while the free-entry number is infinite (see Footnote 26). Hence, free entry obviously leads
to socially excessive entry under both criteria. Furthermore, the level of entry is at least
as high under second-best than under first-best regulation.28 Thus our results substantially
clarify and extend the analysis of Suzumura and Kiyono (1987).
5.3
The Herfindahl Concentration Index and Welfare
The Herfindahl-Hirschman Index (or HHI) of industry concentration, defined as a (normalized) sum of the squares of the firms’ market shares, is the most often used quantitative
assessment of industry concentration. In particular, the value of the HHI constitutes the
primary indicator for antitrust authorities of market power and of the likelihood of overt
or tacit collusion in a given market. The HHI is also one of the main elements of the 1982
Merger Guidelines29 in determining whether a proposed merger is to be allowed.
Underlying the extensive reliance of economic law on this measure is a fundamental
belief that social welfare and the HHI are always inversely related (see e.g. Dansby and
Willig, 1979). Yet, this belief has recently been challenged by theoretical studies based
on the Cournot model. Farrell and Shapiro (1990) establish that, with a fixed number of
(nonidentical) firms, whenever industry output is unchanged folowing individual firm output
26 The
latter follows from the fact that Cournot equilibria are intersections of r with the rays y/(n − 1), and
(i) the fact that r has all its slopes greater than −1, if ∆ > 0, or (ii) the fact that r is continuous if a firm’s
profit is quasi-concave in own output (whether ∆ < 0 globally or not).
27 These may also prevail over the relevant range of operation regardless of the properties of the ”variable”
cost function in the case of costly entry (with the entry cost acting as a fixed cost). Hence, our conclusions
here may also apply when entry is costly.
28 This is actually a new result here, as Suzumura and Kiyono assume convex costs in the comparison between
the free-entry and the first-best outcomes.
29 For some historical backround on these Guidelines and an exchange of views among experts, see the
Symposium in the Journal of Economic Perspectives, vol. 1, 1987.
21
changes, social welfare and the HHI must change in the same direction. Salant and Shaffer
(1999) provide further insight into this result in the case of constant unit costs by showing
that both welfare and the HHI increase if and only the variance of the unit costs increases
in a mean-preserving way. Also see Daughety (1990).
The present paper sheds some further light on this issue by considering the effects of
changing the number of firms instead. Given the symmetry assumption, the HHI with n
firms here is clearly given by (a constant factor of) 1/n. Hence, the HHI decreases if and
only if the number of firms increases. On the other hand, our results indicate that in the
presence of scale economies (with A0 < 0), social welfare decreases if the number of firms
exceeds some socially optimal level n∗ . Thus, both the HHI and welfare decrease whenever
n increases beyond n∗ . In particular, in industries where n∗ = 1, the two measures would
always produce conflicting prescriptions as the number of firms increases.
This conclusion clearly suggests that the HHI should be augmented by some measure
of economies of scale in the industry that would allow appropriate balancing between the
legitimate fears of market power and the desire for production efficiency.
5.4
Concentration and Profitability
One of the most extensive debates in industrial organization has revolved around the alleged positive relationship between market concentration and profits. While the majority
view ended up with a belief that a weak correlation exists, the issue remains somewhat
controversial. In spite of the simplicity of our framework, our results suggest new possible
theoretical explanations of an elementary nature as to why this issue turned out to be so
complex. Consider a two-stage game of free entry followed by Cournot competition amongst
the entrants. First, the possible multiplicity of Cournot equilibria leads to different entry
levels or different equilibria for the same entry level. Either way, profits per-firm will differ.
Second, our results confirm in a general way that scale economies should imply a more
pronounced level of this correlation (Demsetz, 1974, Dewey, 1976 and Lambson, 1987): Since
industry profits increase with concentration, per-firm profits increase at an increasing rate
(i.e. π n ≥
(n+1)
π n+1 ).
n
Conversely, if with diseconomies of scale, we have πn ≤
correlation between profits and concentration is more likely to be weak.30
30
22
(n+1)
π n+1 ,
n
the
Third, with strong scale economies (∆ < 0), the free-entry number of firms is fully indeterminate. In particular, with no entry cost, any number of firms is possible in a subgameperfect equilibrium, thus leading to many possible profit levels! Postulating that the actual
number of firms is determined in part by historical and other random events in such markets,
no clear correlation could be expected between profits and industry characteristics.
5.5
Destructive Competition
Destructive competition was a recurrent theme in older case and empirical studies of regulated industries, particularly those in the transportation sector such as railroad and trucking
(see Sharkey (1982) for a historical account). It is typically associated with a combination
of industry characteristics, such as strong economies of scale (often due to large fixed costs),
large productive capacity, relatively easy entry, and ill-guided government subsidies. The
symptoms of destructive competition in such industries include high levels of market instability, excessive capacity and widespread price discrimination, often leading to frequent
changes in regulatory regimes, including entry regulation.
Sharkey (1982) develops a cooperative game-theoretic approach to model destructive
competition, defining industry stability by the nonemptiness of the core. The results here
suggest a simple and natural alternative within the noncooperative paradigm: In the abscence of any regulatory interference, destructive competition can be fruitfully modelled by
Cournot competition under the assumption that ∆ is globally negative. Indeed, increases
in competition from any pre-existing level, including in particular monopoly, result in lower
consumer welfare, per-firm profit and social welfare. Thus higher competition is unambiguously detrimental to all economic agents, with even unregulated monopoly emerging as the
best among market outcomes. Furthermore, and more strikingly, some aspects of reported
market instability in industries thought to have undergone phases of destructive competition may be instructively linked to the indeterminacy in the number of active firms and the
unstable nature of the Cournot equilibria (in the sense of divergence of best-reply dynamics),
both of which are characteristics of the case ∆ globally negative (see Section 4.4.)
Furthermore, three other important factors of a more dynamic nature contribute to the higher correlation
in the case of economies of scale: Mergers are more likely to be sought and to be allowed, entry by new firms
is more difficult (in particular if fixed costs are sizable), and collusion is thus more likely.
23
5.6
Natural Monopoly
Following various attempts, Baumol, Panzar and Willig (1982) provided the final definition of
natural monopoly: An industry with a subadditive cost function. This is the least restrictive
property of a cost function that captures the notion that any amount of final output is
cheaper to produce by one firm, or, in other words, subdividing production cannot possibly
save on costs. This definition completely ignores the demand side of the market, which is
justified in light of two special features that were dominant in the economic scene two decades
ago. The first, reflecting the prevalent public policy view of the times, is that monopolies
are to be regulated anyway, so that market conduct is not really an issue, leaving production
efficiency as the primary concern. The second, a theoretical belief, is that if an industry
has a downward-sloping average cost curve and the market is contestable, the only stable
configuration will involve a single firm pricing at average cost, resulting in zero profits.
Subsequently, a near-conscensus emerged, recognizing the limited real-life validity of
contestable markets31 , and a wave of deregulation originating in the US and the UK swept
through the industrialized world. In view of the need to incorporate the demand side of the
market now in a revised definition of natural monopoly, the analysis of the present paper
suggests an obvious alternative: An unregulated monopoly is natural if social welfare Wn is
maximized by n = 1. According to our results, this would require scale economies of sufficient
magnitude over the relevant range, but not necessarily that ∆ be globally < 0. Recall that
the Main Example shows that n∗ can be equal to 1 for an industry for which ∆ > 0 globally.
This definition is clearly more restrictive than the old one32 , as it incorporates the market
or demand side of the industry. In other words, it strikes a socially optimal balance between
the detrimental effects of concentration and the cost-saving effects of size.
More generally, a natural n∗ -firm oligopoly can be analogously defined by n∗ = arg maxn Wn .
If ∆ < 0, Proposition 10 implies that n∗ must necessarily be equal to 1. Hence, n∗ > 1 is
not compatible with ∆ < 0 globally. On the other hand, it is compatible with ∆ > 0 and
with ∆ not having a uniform sign on all its domain.
31 In
a book review of Baumol, Panzar and Willig (1982), Spence (1983) offered an eloquent account of the
currently prevailing view on contestable markets. He concludes that the benefits of this theory lie in its
thorough analysis of cost functions, and not in its empirical relevance as a theory of market structure.
32 Scale economies of any magnitude imply the subadditivity of the cost function, but not vice-versa.
24
5.7
Merger Policy
Proposition 3 says that with scale economies, there are industry-wide gains to mergers.
However, according to Cournot theory on mergers, in a unilateral merger, these gains are
generally appropriated by nonparticipating firms, except in near-monopolization cases.33
An important aspect of the 1984 revisions of the Merger Guidelines is their novel consideration of cost efficiency: mergers that are likely to raise prices will be allowed if the
merging firms can demonstrate by ”clear and convincing evidence” that the merger will lead
to significant cost savings or efficiency benefits. In practice, this clause has been exploited by
candidate firms to secure approval by overstating uncertain future cost gains. Yet, ignoring
the important verification and burden-of-proof issues here, our conclusions are certainly in
favor of this amendment to the Guidelines for industries with known scale economies, and
suggest that the extent of scale economies in the industry should be taken into account, as
opposed to more firm-specific claims of technological and organizational synergies.
REFERENCES
Amir, R. (1996), ”Cournot Oligopoly and the Theory of Supermodular Games”, Games
and Economic Behavior, 15, 132-148.
Amir R., and V. E. Lambson (2000), “On the Effects of Entry in Cournot Markets,”
Review of Economic Studies, to appear.
d’Aspremont, C., R. Dos Santos Ferreira, L.-A. Gerard-Varet (2000), Contestability
and the Indeterminacy of Free-Entry Equilibria, Japanese Economic Review, to appear.
Bain, J. (1956), Barriers to New Competition, Harvard University Press, Cambridge.
Baumol, W., J. C. Panzar and R. Willig (1982), ”Contestable Markets and the Theory
of Industry Structure”, Harcourt Brace Jovanovich, New York.
Brozen, Y. (1973), ”Concentration and Profits: Does Concentration Matter?” in J.F.
Weston and S. Ornstein, eds. The Impact of Large Firms on the US Economy, Lexington Books, 59-70.
Cox, J. and M. Walker (1998), ”Learning to Play Cournot Duopoly Strategies”, Journal
of Economic Behavior and Organization, 36, 141-161.
33 See
e.g. Farrell and Shapiro (1990) and Fauli-Oller (1997).
25
Daughety, A. (1990), Beneficial Concentration, American Economic Review, 80, 123137.
de Meza, D. (1985), ”A Stable Cournot-Nash Industry Need not be QuasiCompetitive”, Bulletin of Economic Research, 37, 153, 156.
Demsetz, H. (1973), “Industry Structure, Market Rivalry and Public Policy, Journal
of Law and Economics, 16, 1-10.
Demsetz, H. (1974), “Two Systems of Belief about Monopoly,” in Harvey Goldschmid
et al., eds., Industrial Concentration: The New Learning, Boston: Little, Brown, 164184.
Dewey, D. (1976), “Industrial Concentration and the Rate of Profit: Some Neglected
Theory,” Journal of Law and Economics, 19, 67-78.
Farrell, J. and C. Shapiro (1990), ”Asset Ownership and Market Structure in
Oligopoly”, RAND Journal of Economics, 21, 275-292.
Fauli-Oller, R. (1997), ”On Merger Profitability in a Cournot Setting”, Economics
Letters, 54(1), 75-79.
Frank, C. R. (1965), ”Entry in a Cournot Market”, Review of Economic Studies, 32,
245-250.
Goldschmid, H., H. M. Mann and J.F. Weston (1974), eds., Industrial Concentration:
The New Learning, Boston: Little, Brown.
Fudenberg, D. and J. Tirole (1991), Game Theory, MIT Press: Cambridge, MA.
Hahn, F. (1962), The Stability of the Cournot Oligopoly Solution, Review of Economic
Studies, 29, 329-221.
Holt, C. (1995), ”Industrial Organization: A Survey of the Results of Laboratory
Experiments”, in Handbook of Experimental Economics, A. Roth and J. Kagel eds.,
Princeton University Press.
Lambson, V. E. (1987), “Is the Concentration-Profit Correlation Partly an Artifact of
Lumpy Technology?” American Economic Review, 77, 731-733.
Lopez-Cunat, J. M. (1999), One-Stage and Two-Stage Entry Cournot Equilibria, Investigaciones Economicas, 23(1), 115-28.
MacManus, M. (1962), ”Number and Size in Cournot Equilibrium”, Yorkshire Bulletin
of Economic and Social Research, 14, 14-22.
Mankiw, N.G. and Whinston, M. D. (1986), ”Free Entry and Social Inefficiency”, Rand
Journal of Economics, 17, 48-58.
de Meza, D. (1985), ”A Stable Cournot-Nash Industry Need not be QuasiCompetitive”, Bulletin of Economic Research, 37, 153-156.
26
Milgrom, P. and J. Roberts (1990), ”Rationalizability, Learning, and Equilibrium in
Games with Strategic Complementarities”, Econometrica, 58, 1255-1278.
Milgrom, P. and J. Roberts (1994), ”Comparing Equilibria”, American Economic Review, 84, 441-459.
Mueller, D. (1986), Profts in the Long Run, Cambridge University Press, New-York.
Novshek, W. (1980), ”Cournot Equilibrium with free entry”, Review of Economic
Studies, 47, 473-486.
Novshek, W. (1985), ”On the Existence of Cournot Equilibrium”, Review of Economic
Studies, L II, 85-98.
Okuguchi, K. (1973), ”Quasi-Competitiveness and Cournot Oligopoly”, Review of Economic Studies, 40, 145-148.
Perry, M. (1984), ”Scale Economies, Imperfect Competition and Public Policy”, Journal of Industrial Economics, 32, 313-330.
Roberts, J. and H. Sonnenschein (1976), ”On the Existence of Cournot Equilibrium
without Concave Profit Functions”, Journal of Economic Theory, 13, 112-117.
Robson, A. (1990), ”Stackelberg and Marshall”, American Economic Review, 80, 6982.
Ruffin, R. J. (1971), ”Cournot Oligopoly and Competitive Behavior”, Review of Economic Studies, 38, 493-502.
Salant, S. and G. Shaffer (1999): “Unequal Treatment of Identical Agents in Cournot
Equilibrium”, American Economic Review, 89, 585-604.
Schelling, T. (1960), The Strategy of Conflict, Harvard University Press, Cambridge,
MA.
Seade, J. (1980a), ”On the Effects of Entry”, Econometrica, 48, 479-489.
Sharkey (1982), The Theory of Natural Monopoly, Cambridge University Press, Cambridge, U.K.
Sherer, F.M. (1974), Economies of Scale and Industrial Concentration, in Harvey
Goldschmid et al., eds., Industrial Concentration: The New Learning, Boston: Little, Brown.
Spence, M. (1983), ”Contestable Markets and the Theory of Industry Structure: A
Review Article”, Journal of Economic Literature, 21(3), 981-90.
Suzumura, K. and K. Kiyono (1987), ”Entry Barriers and Economic Welfare”, Review
of Economic Studies, L IV, 157-167.
27
Topkis, D. (1978), ”Minimizing a Submodular Function on a Lattice, Operations Research, 26, 305-321.
von Weiszacker, C. (1980), ”A Welfare Analysis of Barriers to Entry”, Rand Journal
of Economics, 11, 399-420.
6.
Appendix
Here, we provide (A1) a graphical illustration of the benefits of our approach to equilibrium
comparisons when ∆ > 0 and there are multiple equilibria, and (A2) the formal proofs.
(A1) Comparing Multiple Equilibria. The following discussion refers to Figure 1.
With ∆ > 0, the reaction curve r can never decrease at a rate larger than −1. With n firms,
the Cournot equilibria are the intersections of the reaction curve with the line y/(n − 1).
The number of equilibria is thus 1 for n = 1 and n = 4, 3 for n = 2, and 5 for n = 3. It is
easy to see that:
1) For the extremal equilibria, industry output zn increases with n (to see this, simply
draw lines of slope −1 through these equilibria and observe the outward shifts.)
2) The motion of zn is generally indeterminate for the middle equilibria since a may go to
b, c, or d! On the other hand, if there were only one middle equilibrium for both n = 2 and
n = 3, then zn would actually decrease for this unstable equilibrium, so that price would
increase as we go from 2 to 3 firms! More generally, the comparative statics of unstable
equilibria goes opposite that of the extremal equilibria. In particular, price increases with
n. Furthermore, since a firm’s rivals’ total output, yn , decreases with n at an unstable
equilibrium, per-firm profits also increase with n at such equilibria.
3) Most of the other comparisons of interest (industry profit and welfare) depend on the
previous two points, as seen in our proofs. Thus, dealing only with the extremal equilibria,
one avoids a lot of indeterminacy in the comparative statics.
4) Even with the artifact of treating n as a real variable, the implicit function theorem
approach cannot be justified in dealing with (say) dxn /dn (add to this that r need not even
be continuous in general). In addition, signing dxn /dn in hybrid cases (where ∆ is not
uniformly signed) can be misleading (de Meza, 1985).
(A2) Proofs.
Proof of Proposition 3b. Let xn be an extremal Cournot equilibrium, and consider
28
πn
= xn {P (nxn ) − A(xn )}
©
£ n+1
¤
ª
n+1
≥ n+1
x
P
x
+
(n
−
1)x
−
A(
x
)
n+1
n+1
n
n+1
n
n
n
ª
© £ n+1
¤
n+1
n+1
xn+1 )
≥ n xn+1 P n xn+1 + (n − 1) n xn+1 − A( n+1
n
©
ª
n+1
= n+1
x
P
[(n
+
1)x
]
−
A(
x
)
n+1
n+1
n+1
n
n
≥
=
n+1
xn+1
n
n+1
π n+1
n
{P [(n + 1)xn+1 ] − A(xn+1 )}
where the first inequality follows from the Cournot equilibrium property, the second from
the facts that P is decreasing and nxn ≤ (n + 1)xn+1 , and the third from the fact that
xn+1 ) ≤ A(xn+1 ). Multiplying across by n gives the conclusion.
A( n+1
n
The global statement follows directly from the local statement.
Proof of Proposition 4. With z denoting industry output, industry profit is given by
Πn (z) = z[P (z) − A(z/n)]. Since for fixed z, Πn (z) is increasing in n if A0 > 0, the result
follows from the envelope theorem, as monopoly profit π 1 = maxz Π1 (z).
∧
Proof of Proposition 5. Since ∆ > 0 or P 0 (z)−C 00 (z−y) < 0 on ϕ = {(z, y) : y ≥ 0, z ≥ y},
Rz
the function W (z) , 0 P (t)dt − nC(z/n) is concave in z. Now, consider
Wn0 − Wn
=
R zn0
0
Rz
P (t)dt − nC(zn0 /n) − [ 0 n P (t)dt − nC(zn /n)]
= W (zn0 ) − W (zn )
≥ W 0 (zn0 )(zn0 − zn ) , since W is concave.
= [P (zn0 ) − C 0 (zn0 /n)]n(x0n − xn )
≥ 0, since x0n ≥ xn .
The second statement of the lemma follows directly.
Proof of Proposition 6b. To prove Part (b)(i), consider:
Wn+1 − Wn
=
=
©R zn+1
0
R zn+1
zn
ª ©R z
ª
P (t)dt − zn+1 A(xn+1 ) − 0 n P (t)dt − zn A(xn )
P (t)dt − zn+1 A(xn+1 ) + zn A(xn )
≥ (zn+1 − zn )P (zn+1 ) − zn+1 A(xn+1 ) + zn A(xn )
= zn+1 [P (zn+1 ) − A(xn+1 )] − zn [P (zn+1 ) − A(xn )]
≥ zn+1 [P (zn+1 ) − A(xn+1 )] − zn [P (zn+1 ) − A(xn+1 )]
= (zn+1 − zn ) [P (zn+1 ) − A(xn+1 )] ≥ 0,
29
where the first inequality follows from the fact that P (·) is decreasing, the second from the
assumption A(xn ) ≥ A(xn+1 ), while the last follows from the facts that zn+1 ≥ zn (since
∆ > 0) and xn+1 is a symmetric Cournot equilibrium.
To prove Part (b)(ii), we begin with two preliminary observations. First, the function
R nx
Vn (x) , 0 P (t)dt−nC(x) is concave in x for each n since Vn00 (x) = n[nP 0 (nx)−C 00 (x)] < 0,
as a result of ∆ > 0. Second, since zn+1 = (n + 1)xn+1 and P is decreasing,
Z nxn+1
Z zn+1
Z nxn+1
Z zn+1
P (t)dt =
P (t)dt +
P (t)dt ≥
P (t)dt + xn+1 P (zn+1 ).
0
0
nxn+1
(15)
0
Now, consider,
Wn+1 − Wn
ª ©R z
ª
P (t)dt − (n + 1)C(xn+1 ) − 0 n P (t)dt − nC(xn )
©R nx
ª ©R nx
ª
≥ xn+1 P (zn+1 ) − C(xn+1 ) + 0 n+1 P (t)dt − nC(xn+1 ) − 0 n P (t)dt − nC(xn )
=
©R zn+1
0
= π n+1 + Vn (xn+1 ) − Vn (xn )
≥ π n+1 + Vn0 (xn+1 )(xn+1 − xn )
0
= π n+1 + n[P (nxn+1 ) − C (xn+1 )](xn+1 − xn ) ≥ 0,
where the first inequality follows from (14), the second from the concavity of Vn in x, and
0
the last from the facts that P (nxn+1 ) ≥ P [(n + 1)xn+1 ] ≥ C (xn+1 ) and xn+1 ≥ xn .
Proof of Proposition 10. Consider (with zn ≥ zn+1 here, since ∆ < 0):
Wn − Wn+1
=
=
©R zn
0
R zn
zn+1
ª ©R z
ª
P (t)dt − zn A(xn ) − 0 n+1 P (t)dt − zn+1 A(xn+1 )
P (t)dt − zn A(xn ) + zn+1 A(xn+1 )
≥ (zn − zn+1 )P (zn ) − zn A(xn ) + zn+1 A(xn+1 )
= zn [P (zn ) − A(xn )] − zn+1 [P (zn ) − A(xn+1 )]
≥ zn [P (zn ) − A(xn )] − zn+1 [P (zn ) − A(xn )]
= (zn − zn+1 ) [P (zn ) − A(xn )] ≥ 0
where the first inequality follows from the fact that P (·) is decreasing, the second from the
facts that xn ≥ xn+1 and A(·) is nonincreasing (the latter follows since ∆ < 0 requires
concavity of C), and the last from the fact that zn ≥ zn+1 (since ∆ < 0).
30
Bücher des Forschungsschwerpunkts Marktprozeß und Unternehmensentwicklung
Books of the Research Area Market Processes and Corporate Development
(nur im Buchhandel erhältlich/available through bookstores)
Tobias Miarka
Financial Intermediation and Deregulation: A
Critical Analysis of Japanese Bank-FirmRelationships
2000, Physica-Verlag
Jianping Yang
Bankbeziehungen deutscher Unternehmen:
Investitionsverhalten und Risikoanalyse
2000, Deutscher Universitäts-Verlag
Horst Albach, Ulrike Görtzen, Rita Zobel (Hg.)
Information Processing as a Competitive
Advantage of Japanese Firms
1999, edition sigma
Dieter Köster
Wettbewerb in Netzproduktmärkten
1999, Deutscher Universitäts-Verlag
Christian Wey
Marktorganisation durch Standardisierung: Ein
Beitrag zur Neuen Institutionenökonomik des
Marktes
1999, edition sigma
Horst Albach, Meinolf Dierkes, Ariane Berthoin
Antal, Kristina Vaillant (Hg.)
Organisationslernen – institutionelle und
kulturelle Dimensionen
WZB-Jahrbuch 1998
1998, edition sigma
Lars Bergman, Chris Doyle, Jordi Gual, Lars
Hultkrantz, Damien Neven, Lars-Hendrik Röller,
Leonard Waverman
Europe’s Network Industries: Conflicting
Priorities - Telecommunications
Monitoring European Deregulation 1
1998, Centre for Economic Policy Research
Manfred Fleischer
The Inefficiency Trap
Strategy Failure in the
German Machine Tool Industry
1997, edition sigma
Christian Göseke
Information Gathering and Dissemination
The Contribution of JETRO to
Japanese Competitiveness
1997, Deutscher Universitäts-Verlag
Andreas Schmidt
Flugzeughersteller zwischen globalem
Wettbewerb und internationaler Kooperation
Der Einfluß von Organisationsstrukturen auf
die Wettbewerbsfähigkeit von
Hochtechnologie-Unternehmen
1997, edition sigma
Horst Albach, Jim Y. Jin, Christoph Schenk (eds.)
Collusion through Information Sharing?
New Trends in Competition Policy
1996, edition sigma
Stefan O. Georg
Die Leistungsfähigkeit japanischer Banken
Eine Strukturanalyse des Bankensystems in
Japan
1996, edition sigma
Stephanie Rosenkranz
Cooperation for Product Innovation
1996, edition sigma
Horst Albach, Stephanie Rosenkranz (eds.)
Intellectual Property Rights and Global
Competition - Towards a New Synthesis
1995, edition sigma.
David B. Audretsch
Innovation and Industry Evolution
1995, The MIT Press.
Julie Ann Elston
US Tax Reform and Investment: Reality and
Rhetoric in the 1980s
1995, Avebury
Horst Albach
The Transformation of Firms and Markets:
A Network Approach to Economic
Transformation Processes in East Germany
Acta Universitatis Upsaliensis, Studia Oeconomiae
Negotiorum, Vol. 34
1994, Almqvist & Wiksell International
(Stockholm).
Horst Albach
"Culture and Technical Innovation: A CrossCultural Analysis and Policy
Recommendations"
Akademie der Wissenschaften zu Berlin (Hg.)
Forschungsbericht 9, S. 1-597
1994, Walter de Gruyter.
DISCUSSION PAPERS 1999
Suchan Chae
Paul Heidhues
Bargaining Power of a Coalition in Parallel Bargaining:
Advantage of Multiple Cable System Operators
Christian Wey
Compatibility Investments in Duopoly with Demand
FS IV 99 - 2
Side Spillovers under Different Degrees of Cooperation
Horst Albach
FS IV 99 - 1
Des paysages florissants? Une contribution
à la recherche sur la transformation
FS IV 99 - 3
The Development of British Competition Law:
A Complete Overhaul and Harmonization
FS IV 99 - 4
Union Power and Product Market Competition:
Evidence from the Airline Industry
FS IV 99 - 5
Justus Haucap
Uwe Pauly
Christian Wey
The Incentives of Employers’ Associations to
Raise Rivals’ Costs in the Presence of
Collective Bargaining
FS IV 99 - 6
Jianbo Zhang
Zhentang Zhang
Asymptotic Efficiency in Stackelberg Markets
with Incomplete Information
FS IV 99 - 7
Standortwahl als Franchisingproblem
FS IV 99 - 8
A Comparison of Multiple-Unit All-Pay and
Winner-Pay Auctions Under Incomplete
Information
FS IV 99 - 9
Collusion with Private and Aggregate Information
FS IV 99 - 10
Strategic Information Revelation and Revenue Sharing
in an R&D Race with Learning Labs
FS IV 99 - 11
Incomplete Information in the Samaritan's Dilemma:
The Dilemma (Almost) Vanishes
FS IV 99 - 12
Market Integration and Market Structure in the
European Soft Drinks Industry: Always Coca-Cola?
FS IV 99 - 13
Pinelopi Koujianou Goldberg
Frank Verboven
The Evolution of Price Discrimination in the
European Car Market
FS IV 99 - 14
Olivier Cadot
Lars-Hendrik Röller
Andreas Stephan
A Political Economy Model of Infrastructure
Allocation: An Empirical Assessment
FS IV 99 - 15
Industriestandort mit Vorbildfunktion?
Das ostdeutsche Chemiedreieck
FS IV 99 - 16
Testing Dynamic Oligopolistic Interaction:
Evidence from the Semiconductor Industry
FS IV 99 - 17
Jeremy Lever
Damien J. Neven
Lars-Hendrik Röller
Zhentang Zhang
Justus Haucap
Christian Wey
Yasar Barut
Dan Kovenock
Charles Noussair
Jim Y. Jin
Jos Jansen
Johan Lagerlöf
Catherine Matraves
Holger Derlien
Tobias Faupel
Christian Nieters
Christine Zulehner
Johan Lagerlöf
Costly Information Acquisition and
Delegation to a “Liberal” Central Banker
FS IV 99 - 18
Ralph Siebert
New Product Introduction by Incumbent Firms
FS IV 99 - 19
Ralph Siebert
Credible Vertical Preemption
FS IV 99 - 20
Ralph Siebert
Multiproduct Competition, Learning by Doing
and Price-Cost Margins over the Product Life Cycle:
Evidence from the DRAM Industry
FS IV 99 - 21
Michael Tröge
Asymmetric Information Acquisition in Credit Auction
FS IV 99 - 22
Michael Tröge
The Structure of the Banking Sector, Credit
Screening and Firm Risk
FS IV 99 - 23
Michael Tröge
Monitored Finance, Usury and Credit Rationing
FS IV 99 - 24
Multimarket Contact, Collusion and the
International Structure of Firms
FS IV 99 - 25
Dokumentation der „Bonner Stichprobe“ – Zur
Datenbank der Jahresabschlüsse deutscher
Aktiengesellschaften, 1960-1997
FS IV 99 - 26
Tomaso Duso
Endogenous Switsching Costs and the Incentive
for High Quality Entry
FS IV 99 - 29
Jos Jansen
Regulating Complementary Input Supply:
Production Cost Correlation and Limited Liability
FS IV 99 - 30
Suchan Chae
Paul Heidhues
The Effects of Downstream Distributor Chains on
Upstream Producer Entry: A Bargaining Perspective
FS IV 99 - 35
Tobias Miarka
The Recent Economic Role of Bank-FirmRelationships in Japan
FS IV 99 - 36
Demand for Customized Products, Production
Flexibility, and Price Competition
FS IV 99 - 37
Silke Neubauer
Horst Albach
Thomas Brandt
Holger Jakob
M. A. Paradowska-Thimm
Jianping Yang
William Novshek
Lynda Thoman
DISCUSSION PAPERS 2000
Justus Haucap
Uwe Pauly
Christian Wey
Collective Wage Setting When Wages Are
Generally Binding: An Antitrust Perspective
FS IV 00 - 01
Regionale Infrastrukturpolitik und ihre Auswirkung
auf die Produktivität: Ein Vergleich von
Deutschland und Frankreich
FS IV 00 - 02
Political Economy of Infrastructure Investment
Allocation: Evidence from a Panel of
Large German Cities
FS IV 00 - 03
Andreas Blume
Asher Tishler
Security Needs and the Performance
of the Defense Industry
FS IV 00 - 04
Tomaso Duso
Who Decides to Regulate? Lobbying
Activity in the U.S. Cellular Industry
FS IV 00 - 05
Hiding Information in Electoral Competition
FS IV 00 - 06
Stephanie Aubert
Andreas Stephan
Achim Kemmerling
Andreas Stephan
Paul Heidhues
Johan Lagerlöf
Absender/From:
Versandstelle - WZB
Reichpietschufer 50
D-10785 Berlin
BESTELLSCHEIN / ORDERFORM
Bitte schicken Sie mir aus der Liste der
Institutsveröffentlichungen folgende Papiere zu:
Bitte schicken Sie bei Ihren Bestellungen von WZB-Papers
unbedingt eine 1-DM-Briefmarke pro paper und einen an
Sie adressierten Aufkleber mit. Danke.
For each paper you order please send a "CouponRéponse International" (international money order)
plus a self-addressed adhesive label. Thank You.
Please send me the following papers from your Publication List:
Paper Nr./No.
Autor/Author + Kurztitel/Short Title